[Patch, fortran] PR51976 - [F2003] Support deferred-length character components of derived types (allocatable string length)

[W Spector]

Janus,

The 'No initializer for component...' error, when the derived type 
definition is in a module, is really a problem.  It seriously restricts 
the usefulness of the feature for obvious reasons.

Also two more issues:

4.) Array assignment doesn't work.  This is mentioned in the revised web 
page, but the compiler accepts the code without indicating an 
unsupported case and then does nothing:

wws@w6ws-4:~/fortran/strings$ cat deflen_test6.F90
program deflen_test6
   implicit none

   type def_strings
#if !defined (FIXED_LENGTH)
     character(:), allocatable :: strings(:)
#else
     character(16), allocatable :: strings(:)
#endif
   end type

   type(def_strings) :: my_strings

   my_strings%strings = [  &
     "first string ",  &
     "second string",  &
     "third string "  &
   ]

   print *,  my_strings%strings

end program
wws@w6ws-4:~/fortran/strings$ /usr/local/gcc-trunk/bin/gfortran 
deflen_test6.F90
wws@w6ws-4:~/fortran/strings$ a.out

wws@w6ws-4:~/fortran/strings$
wws@w6ws-4:~/fortran/strings$ /usr/local/gcc-trunk/bin/gfortran 
-DFIXED_LENGTH deflen_test6.F90
wws@w6ws-4:~/fortran/strings$ a.out
  first string    second string   third string
wws@w6ws-4:~/fortran/strings$


5.) This is a problem with the existing non-derived type capability. 
(It is also a problem in my 4.7.2 compiler.)  Probably should be the 
subject of a different PR:

wws@w6ws-4:~/fortran/strings$ cat deflen_test7.f90
program deflen_test7
   implicit none

   character(:), allocatable :: localstrings(:)

   localstrings = [  &
     "first string ",  &
     "second string",  &
     "third string "  &
   ]
   print *, localstrings

end program
wws@w6ws-4:~/fortran/strings$ /usr/local/gcc-trunk/bin/gfortran 
deflen_test7.f90
wws@w6ws-4:~/fortran/strings$ a.out
  third string third string third string
wws@w6ws-4:~/fortran/strings$

Walter


On 03/09/2014 08:38 AM, W Spector wrote:
>> I have just committed the patch as r208386, thereby implementing
>> deferred-length character components on 4.9 trunk. One big plea to the
>> users: Please test this as soon as possible!
>
> Janus - I have updated my test compiler to the current trunk.  A couple
> of things:
>
> 1.) I can reproduce the 'gimplification' ICE in comment #8 with the
> ALLOCATE statement,
>
> 2.) I can reproduce the 'No initializer for component...' error also
> mentioned in comment #8,
>
> 3.) My example in comment #12 appears to work fine.  However when
> running the example under valgrind, memory leaks are indicated.  I cut
> the example down quite a bit, and it seems that there are leaks
> regardless of whether derived types are used or not.  So this isn't a
> regression due to this patch, but something that has been around for a
> while:
>
> wws@w6ws-4:~/fortran/strings$ cat deflen_test5.f90
> program strings
>    implicit none
>
>    character(:), allocatable :: string
>
>    string = "first string"
>    print *, string
>    string = "second string"
>    print *, string
>    string = "3rd string"
>    print *, string
>
> end program
> wws@w6ws-4:~/fortran/strings$ /usr/local/gcc-trunk/bin/gfortran
> --version deflen_test5.f90
> GNU Fortran (GCC) 4.9.0 20140306 (experimental)
> Copyright (C) 2014 Free Software Foundation, Inc.
>
> GNU Fortran comes with NO WARRANTY, to the extent permitted by law.
> You may redistribute copies of GNU Fortran
> under the terms of the GNU General Public License.
> For more information about these matters, see the file named COPYING
>
> wws@w6ws-4:~/fortran/strings$ valgrind --leak-check=full a.out
> ==3153== Memcheck, a memory error detector
> ==3153== Copyright (C) 2002-2011, and GNU GPL'd, by Julian Seward et al.
> ==3153== Using Valgrind-3.7.0 and LibVEX; rerun with -h for copyright info
> ==3153== Command: a.out
> ==3153==
>   first string
>   second string
>   3rd string
> ==3153==
> ==3153== HEAP SUMMARY:
> ==3153==     in use at exit: 10 bytes in 1 blocks
> ==3153==   total heap usage: 24 allocs, 23 frees, 11,851 bytes allocated
> ==3153==
> ==3153== 10 bytes in 1 blocks are definitely lost in loss record 1 of 1
> ==3153==    at 0x4C2B4F0: realloc (in
> /usr/lib/valgrind/vgpreload_memcheck-amd64-linux.so)
> ==3153==    by 0x400B0E: MAIN__ (in /home/wws/fortran/strings/a.out)
> ==3153==    by 0x400C12: main (in /home/wws/fortran/strings/a.out)
> ==3153==
> ==3153== LEAK SUMMARY:
> ==3153==    definitely lost: 10 bytes in 1 blocks
> ==3153==    indirectly lost: 0 bytes in 0 blocks
> ==3153==      possibly lost: 0 bytes in 0 blocks
> ==3153==    still reachable: 0 bytes in 0 blocks
> ==3153==         suppressed: 0 bytes in 0 blocks
> ==3153==
> ==3153== For counts of detected and suppressed errors, rerun with: -v
> ==3153== ERROR SUMMARY: 1 errors from 1 contexts (suppressed: 2 from 2)
> wws@w6ws-4:~/fortran/strings$
>
>
> Walter
>